Download Chapter 1. Linear equations

Chapter 1. Linear equations
Review of matrix theory
Fields
System of linear equations
Row-reduced echelon form
Invertible matrices
Fields
• Field F, +, 
F is a set. +:FxFF, :FxFF
– x+y = y+x, x+(y+z)=(x+y)+z
–
unique 0 in F s.t. x+0=x
–
unique -x s.t. x+(-x) = 0
– xy=yx, x(yz)= (xy)z
–
unique 1 in F s.t. x1 = x
–
unique
s.t.
– x(y+z) = xy+yz
• A field can be thought of as a generalization of the field of real
numbers useful for some other purposes which has all the
important properties of real numbers.
• To verify something is a field, we need to
show that the axioms are satisfied.
–
–
–
–
The real number field R
Complex number field
The field of rational numbers Q
The set of natural numbers N is not a field.
• For example 2x z = 1 for no z in N. (no -x also.)
– The set of real valued 2x2 matrices is not a field.
• For example
for no A.
• Consider: Zp ={ 0, 1, 2, …, p-1}
– For p =5, 9=4 mod 5. 1+4 = 0 mod 5.
3 4 = 2 mod 5. 3 2 = 1 mod 5.
– If p is not prime, then the above is not a field. For
example, let p=6. 2.3= 0 mod 6. If 2.x = 1 mod 6,
then 3=1.3=2.x.3=2.3.x=0.x=0.
A contradiction.
– If p is a prime, like 2,3,5,…, then it is a field. The
proof follows:
Zp ={ 0, 1, 2, …, p-1} is a field if p is a prime number
0 and 1 are obvious. For each x, -x equals p-x.
Thus a’ is the inverse of x.
Other axioms are easy to verify by following
remainder rules well.
In fact, only the multiplicative inverse axiom
fails if p is not a prime.
Characteristic
• A characteristic of a field F is the
smallest natural number p such that
p.1=1+…+1 = 0.
• If no p exists, then the characteristic is
defined to 0.
• p is always a prime or 0. (r, s natural
number
If (rs)1=0, then by distributivity r1.s1=0,=> r1=0 or
s1=0)
• p.x = 0 for all x in F.
• For R, Q, the chars are zero. p for Zp
• A subfield F’ of a field F is a subset
where F’ contains 0, 1, and the
operations preserve F’ and inverses are
in F’.
– Example:
– A subfield F’’ of a subfield F’ of a field F is
a subfield of F.
A system of linear equations
• Solve for
A11 x1 + A12 x2 + + A1n xn = y1
A21 x1 + A22 x2 + + A2n xn = y2
Am1 x1 + Am2 x2 +
+ Amn xn = ym
– This is homogeneous if
– To solve we change to easier problem by row operations.
Elementary row operations
– Multiplication of one row of A by a scalar in F-{0}.
– Replacement of r th row of A by row r plus c times
s th row of A (c in F, r  s)
– Interchanging two rows
• An inverse operation of elementary row
operation is a row operation,
• Two matrices A, B are row-equivalent if one
can make A into B by a series of elementary
row operations. (This is an equivalence relation)
• Theorem: A, B row-equivalent mxn
matrices. AX=0 and BX=0 have the
exactly same solutions.
• Definition: mxn matrix R is row-reduced if
– The first nonzero entry in each non-zero row of R is 1.
– Each column of R which contains the leading non-zero entry
of some row has all its other entries 0
• Definition: R is a row-reduced echelon matrix if
– R is row-reduced
– Zero rows of R lie below all the nonzero rows
– Leading nonzero entry
of row i:
(r  n since strictly increasing)
• The main point is to use the first nonzero
entry of the rows to eliminate entries in the
column. Sometimes, we need to exchange
rows. This is algorithmic.
• In this example:
• Theorem: Every mxn matrix A is rowequivalent to a row-reduced echelon form.
• Analysis of RX=0. R mxn matrix
– Let r be the number of nonzero rows of R. Then r n
– Take Variables of X:
– Remaining variables of X:
– RX=0 becomes
– All the solutions are obtained by assigning any
values to
– If r < n, n-r is the dimension of the solution space.
– If r = n, then only X=O is the solution.
• Theorem 6: A mxn m< n. Then AX=0 has a nontrivial
solution.
• Proof:
– R r-r-e matrix of A.
– AX=0 and RX=0 have same solutions.
– Let r be the number of nonzero rows of R.
– r  m < n.
• Theorem 7. A nxn. A is row-equivalent to I iff AX=0
has only trivial solutions.
• Proof:  AX=0, IX=0 have same solutions.
–
–
–
–
–
 AX=0 has only trivial solutions. So does RX=0.
Let r be the no of nonzero rows of R.
rn since RX=0 has only trivial solutions.
But rn always. Thus r=n.
R has leading 1 at each row. R = I.
• Matrix multiplications
• A(BC) = (AB)C A: mxn B:nxr C:rxk
• Elementary matrix E (nxn) is obtained from I by a
single elementary move.
• Theorem 9: e elementary row-operation
E mxm elementary matrix E = e(I). Then
e(A)=E.A= e(I).A.
• Corollary: A, B mxn matrices.
B is row-equivalent to A iff B=PA where
P is a product of elementary matrices
Invertible matrices
• A nxn matrix.
–
–
–
–
If BA = I B nxn, then B is a left inverse of A.
If AC=I C nxn, then C is a right inverse of A
B s.t. BA=I=AB. B is the inverse of A
We will show finally, these notions are equivalent.
• Lemma: If A has a left inverse B and a right
inverse C, then B = C.
– Proof:B=BI=B(AC)=(BA)C=IC=C.
• Theorem: A, B nxn matrices.
– (i) If A is invertible, so is A-1. (A-1)-1=A.
– (ii) If both A,B are invertible, so is AB and
(AB)-1=B-1A-1.
– Products of invertible matrices are
invertible.
• Theorem: An elementary matrix is
invertible. e an operation, e1 inverse
operation. Let E = e(I). E1=e1(I). Then
EE1=e(E1)= e(e1(I))=I. E1E=e1(e(I))=I.
• Theorem 12: A nxn matrix. TFAE:
– (I) A is invertible.
– (ii) A is row-equivalent to I.
– (iii) A is a product of elementary matrices.
•
proof:
–
–
–
–
Let R be the row reduced echelon matrix of A.
R=Ek…E1A. A= E1-1…Ek-1R.
A is inv iff R is inv.
R is inv iff R=I
• (() if RI. Then exists 0 rows.
R is not inv.() R=I is invertible. )
• Fact: R = I iff R has no zero rows.
• Corollary: A I by a series of row
operations. Then I A-1 by the same
series of operations.
– Proof:
• I = Ek…E1A.
• By multiplying both sides by A-1 .
• A-1= Ek…E1. Thus, A-1= Ek…E1I.
• Corollary: A,B mxn matrices
B is row-equivalent to A iff B=PA for an
invertible mxm matrix P.
• Theorem 13: A nxn TFAE
– (i) A is invertible
– (ii) AX=O has only trivial solution.
– (iii) AX=Y has a unique solution for each nx1
matrix Y.
• Proof: By Theorem 7, (ii) iff A is row-equiv. to
I. Thus, (i) iff (ii).
– (ii) iff (iii)  A is invertible. AX=Y. Solution X=A-1Y.
•  Let R be r-r-e of A. We show R=I.
– We show that the last row of R is not O.
– Let E=(0,0,..,1) nx1 column matrix.
– If RX=E is solvable, then the last row of R
is not O.
– R=PAA=P-1R.
– RX=E iff AX=P-1E which is always solvable
by the assumption (iii).
• Corollary: nxn matrix A with either a left
or a right inverse is invertible.
• Proof:
– Suppose A has a left inverse.
•
• AX=0 has only trivial solutions. By Th 13, done.
– BAX=0 -> X=0.
– Suppose A has a right inverse.
• C has a left-inverse A.
• C is invertible by the first part. C-1=A.
• A is invertible since C-1 is invertible.